Re: Benchmarking for SVN?

I had a benchmarking framework a while I used to compare SVN vs P4 vs CVS.

http://subversion.kicks-ass.org/perfnew/elapsed.html

The hard part about making a benchmark that runs over an extended period
of time is that the results aren't very useful if the machine you are
running doesn't get locked down. Things like the kernel, filesystem
free disk space, etc all affect performance and may skew the results
beyond usefullness.

Maybe a slightly better approach would be to do something in the line of
a CMake like build status page that has these benchmarks built in.
While not very useful by themselves, you can detect problems if suddenly
one platform gets much slower than the others by looking at the averages.

http://www.cmake.org/Testing/Dashboard/20050621-0100-Nightly/Dashboard.html

Dan Notestein wrote:

> Maybe this already exists, but if not, I wanted to suggest that the
> Subversion project adds a benchmark report on a test svn respository
> of a set size and hardware configuration. This would be a nice way to
> measure the performance of svn as it goes through changes, and it
> would also give users a way to recognize if their performance was
> below average due to a bad configuration of some kind. The benchmark
> could include the speeds for the most common commands and set up to
> run automatically on new releases.
>

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Tue Jun 21 08:08:37 2005